van der Tweel I, van Noord P A, Kaaks R
Centre for Biostatistics, University of Utrecht, The Netherlands.
J Clin Epidemiol. 1993 Mar;46(3):253-9. doi: 10.1016/0895-4356(93)90073-a.
Application of sequential analysis may avoid unnecessary experimentation and achieve economical use of available biomaterial stored in biological banks. When, as often happens in cohort case-controls studies, cases are scarce, it may be possible to use multiple control observations per case to increase the power of a test for detecting differences between cases and controls. Samples from a biological data bank were analysed. We compared results of a non-sequential analysis with results of sequential t-tests for 1 to 5 controls matched per case in a cohort nested case-control study. Simulations are performed to get an idea of the unreliability and the power of the sequential test. In general the sequential t-tests are too conservative with respect to the achieved power. Average sample numbers are lower for the sequential tests and decrease with multiple controls. More than 3 or 4 controls per case does not give a meaningful increase in efficiency.
序贯分析的应用可以避免不必要的实验，并实现对生物样本库中现有生物材料的经济利用。在队列病例对照研究中，病例往往较少，此时可以对每个病例使用多个对照观察值，以提高检测病例与对照之间差异的检验效能。我们分析了来自生物数据库的样本。在一项队列巢式病例对照研究中，我们将非序贯分析的结果与针对每个病例匹配1至5个对照的序贯t检验结果进行了比较。通过模拟来了解序贯检验的不可靠性和效能。总体而言，序贯t检验在效能方面过于保守。序贯检验的平均样本数较低，并且随着多个对照的增加而减少。每个病例超过3或4个对照并不会显著提高效率。
